Diversification, Productivity, and Financial Constraints:
Empirical Evidence from the US Electric Utility Industry
Mika Goto
Central Research Institute of Electric Power Industry
Angie Low
Nanyang Business School, Nanyang Technological University
Anil K. Makhija*
Fisher College of Business, The Ohio State University
This Version: February 28, 2008
________________________________________________________________________________
Abstract
We examine the real effects of parent firm diversification on their electric utility operating
companies over the period, 1990-2003.
Since electric utility operating companies produce a single
homogenous product, we can better measure their Total Factor Productivity and make valid
comparisons of productivity across firms. We find that, consistent with a diversification discount,
greater parent diversification is associated with lower productivity across electric utility operating
companies. However, the productivity of the electric utility operating companies improves with
greater parent diversification over time.
Diversification appears to provide an alternative channel
to divert investment dollars away from overinvestment in the core electric business. Finally, we find
that the improvement in the productivity of the electric utility operating companies from greater
parent firm diversification over time is limited to financially constrained firms.
This suggests that
when managers have no resources to waste, it is more likely that any diversification activities are
carefully planned and undertaken for strategic purposes that can help to increase productivity of the
core business.
Key words: Diversification, Total Factor Productivity, Financial Constraints, Electric Utilities
JEL Classification Code: L25, L94
________________________________________________________________________________
*
Corresponding author: Anil K. Makhija, Rismiller Professor of Finance, Fisher College of Business,
The Ohio State University, 2100 Neil Avenue, Columbus, OH 43210. E-mail:
makhija_1@fisher.osu.edu.
1. Introduction
There is a vast literature on how diversification affects the market values of firms.
Following
Lang and Stulz (1994) and Berger and Ofek (1995), many studies have documented that diversified
firms sell at a discount relative to the sum of the values of their stand-alone component segments.
Conglomerates earn lower stock market returns, according to Comment and Jarrell (1985).
implication is that diversification destroys corporate value.
The
However, Campa and Kedia (2002) and
Villalonga (2004) argue that the diversification discount arises endogenously.
They find that
diversifying firms tend to trade at a discount even before they diversify, while Graham, Lemmon, and
Wolf (2002) document that conglomerates tend to purchase already discounted target firms.
Confounding the issue further, Villalonga (2004) points out that the conglomerate discount is a result of
data biases in the Compustat segment data, which are commonly used in research on the diversification
discount. Given this controversy surrounding how diversification affects market valuations, a strand of
research has begun a new approach by examining the real effects of diversification. In a limited
literature so far, Schoar (2002) and Maksimovic and Phillips (2002) have both used plant-level data from
the U.S. Census Bureau on manufacturing firms to study how diversification affects plant productivity.
In this paper, we extend this line of inquiry by examining how the productivity of U. S. electric utility
operating companies was affected by the diversification decisions of their parent firms during the period,
1990 to 2003.
We also examine how the parent’s financial constraints affect the relationship between
its diversification and the productivity of its electric utility operating company. This financing aspect
has not been addressed before.
There are several reasons to study the real effects of diversification on electric utility operating
companies, going beyond the contradictory findings based only on manufacturing firms with additional
evidence from another important industry.
While Schoar (2002) reports that more diversified firms
have higher productivity, according to Maksimovic and Phillips (2002) conglomerate firms are less
productive than single-segment firms because of the significantly lower productivity of peripheral
divisions relative to the main divisions. The electric utility industry offers fertile ground for additional
-1-
evidence because it has engaged in substantial amounts of diversification (57% of all utilities were
engaged in non-electric businesses by 1997, Jandik and Makhija, 2005).
The greater issue here,
however, is the manner in which earlier studies estimate productivity, the main criterion for assessing the
impact of diversification.
As Schoar (2002) herself acknowledges, their estimate of Total Factor
Productivity (TFP) is determined by approximating output by the value of total shipments and changes
in value of inventory.
Thus, their TFP reflects not only the desired variations in efficiency but also
differences in markups.
The problem, of course, lies in the difficulty in formulating physical output
across a sample of manufacturing firms with heterogeneous products. The heterogeneity of products
also implies that their TFP measures are not truly comparable across the firms. In contrast, electric
utility operating companies produce a single, homogenous product, measured in megawatt-hours of
electricity.
Even as we make a case for reexamining the impact of diversification on productivity, we do not
hypothesize a specific impact.
Instead, we argue that the impact of parent diversification on the
productivity of the operating company is an empirical issue.
It has been claimed that diversification
has adverse effects because internal capital markets allocate capital sub-optimally across divisions
(Rajan, Servaes, and Zingales, 2000, Scharfstein and stein, 2000, Scharfstein, 1998, and Shin and Stulz,
1998). On the other hand, Alchian (1969), Weston (1970), Williamson (1975), Gertner, Scharfstein,
and Stein (1994), and Stein (1996) stress the benefits and positive impact of internal capital markets.
Since the electric utility operating company is invariably the core and major business of the parent
firm, the electric utility industry also presents a suitable setting to assess the impact of the parent’s
financial condition on how its diversification affects the productivity of its operating company.
financial
condition
of
the
parent
firm
determines
the
“financial
slack”
to
If the
underwrite
productivity-enhancing investments by the electric utility operating company, we expect that for
financially unconstrained parent firms the relation between parent diversification and operating company
productivity will at least not be adversely affected.
diversification-productivity relationship.
However, financial slack can also worsen the
When managers have relatively more resources than
-2-
investment opportunities, the diversification activities are likely to be “pet projects” that distract
managers from their core business, leading to reductions in productivity. This is a relevant concern for
the electric utility industry.
Many parent firms in the industry historically can be characterized as
having low growth and high free cash flows. These are just the type of firms in which Jensen (1986)
has argued that managers tend to overinvest in self-serving negative net present value projects. This
raises the possibility that diversifying activities were undertaken by many utility managers for
empire-building and entrenchment purposes, all of which take away attention from the main business of
the firm. Indeed, diversification is often argued to be a result of overinvestment (Morck, Shleifer, and
Vishny (1990)).
Acquisitions by cash-rich, low-growth firms perform worse than those of other
acquirers (Lang, Stulz, and Walkling (1991)).
Similarly, Denis, Denis, and Sarin (1997) find that
managerial agency problems explain why firms maintain value-destroying diversification strategies.
Consistent with the agency cost of free cash flow, Harford (1999) finds that cash-rich firms are more
likely to undertake diversifying acquisitions. Notably, in his review of the investments literature, Stein
(2003) stresses that value-destroying over-investments occur only when the level of free cash flow
relative to investment opportunities is greater than expected. Similarly, there are competing explanations
on how diversification matters for the financially constrained firm.
Greater diversification by a
financially constrained firm diverts available dollars from productivity-enhancing investments. On the
other hand, when managers are constrained and have no extra resources to waste, it is likely that any
diversification activities are undertaken for strategic purposes and therefore help to increase productivity
at the core business.
Moreover, when managers are constrained, they often have to go to the external
capital markets for funding and this extra monitoring from the capital markets prevents managers from
engaging in value-destroying activities. This tri-lateral relation between productivity, diversification, and
financial conditions has been ignored in previous work on the real effects of diversification.
We form a matched panel dataset of electric operating companies and their corresponding parent
companies for the 1990-2003 the period.
We then relate the productivity of the operating companies to
diversification activities and financial conditions at the parent company.
-3-
First, we examine the
relationship between operating company TFP and a number of measures of parent diversification
(number of segments, sales herfindahl, asset herfindahl, and fraction of non-utility sales).
We study the
cross-sectional (Fama-MacBeth estimations) and time-series (operating company fixed effects)
relationships. Contrary to Schoar (2002), we find that in the cross-section, more diversified parents are
associated with less productive operating companies, but, in the time series, diversifying activities
increase the productivity of the core electric utility segment. The cross-sectional findings support the
diversification discount reported by Lang and Stulz (1994) and Berger and Ofek (1995). Our results are
also consistent with Maksimovic and Phillips (2002) who use plant-level data and find that diversified
firms are less productive than single-segment firms.
They argue that when firms have relative
comparative advantage across industries, it might be optimal for less efficient firms to diversify. As for
the beneficial impact of diversification over time, it is consistent with the explanation offered by Jandik
and Makhija (2005). Diversification has provided cash-rich utilities a channel for diverting investment
dollars that would have otherwise led to unproductive over-investment in the core electric business.
Next, we examine how the parent firm’s financial condition affects the productivity impact of
diversification.
To take into account both the financial conditions and growth prospects, we make use
of the KZ-index (Kaplan and Zingales, 1997), which measures the degree of financial constraints faced
by a firm, taking into account the cash flows generated, cash on hand, leverage ratios, dividend
payments, and growth opportunities.
Our findings are unchanged when we use the coverage ratio
instead of KZ. Consistent with Jensen’s (1986) agency costs of free cash flow, we find that
diversification undertaken when the parent faces financial slack has a negative impact on productivity at
the operating company.
However, diversification undertaken when the parent company is financially
constrained has a positive impact on the electric segment productivity. These results indicate that
financially constrained firms undertake diversifications more carefully, such that they have a beneficial
effect on the core business.
Further, when managers are constrained, they often have to go to the
external capital markets for funding and this extra monitoring from the capital markets prevents
managers from engaging in productivity-destroying activities.
-4-
The remaining structure of this paper is organized as follows: Section 2 briefly reviews the
literature on diversification and describes the deregulation activities in the utility industry. Section 3
describes the data, the variables used, and the methodology. The results are presented in Section 4.
Section 5 concludes and discusses future agendas.
2. Literature Review on Diversification and the Regulatory Background
2.1.
Literature Review
Several papers have documented that conglomerates trade at a discount relative to single-segment
firms in the same industry. They do this by comparing the market value of conglomerates to the value
of a portfolio of focused firms operating in the same industries as the conglomerate’s divisions. Using
this approach, Lang and Stulz (1994) find that diversified firms have lower values of Tobin’s Q
compared to single-segment firms.
In another study, Berger and Ofek (1995) find that U.S.
conglomerates trade at a discount of 15%. Other papers have confirmed this finding using different
sample periods and different countries.
For example, Servaes (1996) finds a discount for
conglomerates during the 1960s, Lins and Servaes (1999, 2002) document significant discounts in Japan,
the United Kingdom, and a sample of firms from seven emerging markets.1
However, recent literature contests the interpretation that diversification destroys firm value.
Instead, papers such as Campa and Kedia (2002) and Villalonga (2004) argue that the diversification
discounts arises endogenously.
They find that diversifying firms tend to trade at a discount even before
they diversify, while Graham, Lemmon, and Wolf (2002) document that conglomerates tend to purchase
already discounted target firms.
Maksimovic and Phillips (2002) propose a neo-classical,
profit-maximizing model where firms optimally choose the number of segments they operate in
depending on their comparative advantage.
Other studies have also contested the validity of the
conglomerate discount, arguing that Compustat segment data is biased towards finding a conglomerate
1
However, Matsasuka (1993) documents gains to diversifying acquisitions during the 1960s, and
Khanna and Palepu (2000) and Fauver, Houston, and Naranjo (1998) do not find evidence of discounts
in emerging markets.
-5-
discount (see e.g., Villalonga (2004)).
Pertinent to our study, a few papers study the real effects of diversification as an alternative
approach to simply looking at market valuations. Using plant-level data from the U.S. Census Bureau,
Schoar (2002) documents that diversified firms are on average more productive than focused firms.
However, productivity at incumbent plants falls when firms diversify. This is mainly attributed to neglect,
as management turns its attention to the newly-acquired division. Using the same source of data,
Maksimovic and Phillips (2002) find that conglomerates are less productive, especially in their
peripheral segments, consistent with the model of profit maximization they propose. They also find
that firms allocate resources efficiently across the different industry segments.
In examining changes in
productivity during asset reallocations, Maksimovic and Phillips (2001) find that on average,
productivity gains accompany such transactions.
However, the gains depend on the productivity of the
acquiring firm and whether it is the main or peripheral division. The authors, however, cannot rule out
agency considerations that may drive some transactions.
The studies on the real effects of diversification have the benefit of a large sample of
manufacturing firms.
However, the manufacturing industry is very diverse and comparisons of
productivity across firms are problematic given that the firms have very different products.
Consequently, in this study we examine firms producing a single, homogenous product – electricity.
By studying the electric utility industry only, we are able to control for the product heterogeneity,
therefore allowing for a more precise measurement and comparison of productivity across firms.
Furthermore, most studies on diversification have ignored the electric utility industry, primarily because
of the heavy regulation in the industry.
However, deregulation efforts during the 1980s and early 1990s
imply that for most of the 1990s, the managers in the electric utility industry did have considerable
discretion in their diversification and investment projects, similar to firms in other industries.
At any
rate, as a robustness test we take into account the ease of state-level diversification policies.
Since the core business of the electric utilities is the same, we are able to effectively control for
the industry conditions and examine how diversification into non-related businesses affects the
-6-
productivity in the incumbent segments.
Furthermore, we seek to understand under what conditions
would diversification be beneficial and when it would not. Specifically, we concentrate on the motives
managers may have when they diversify. An influential view in the literature is that there are conflicts
of interest between managers and shareholders and that can result in investment decisions being taken
for the private benefits of managers (Jensen and Meckling (1976), Jensen (1986, 1993)), including
diversifying activities that are undertaken for empire-building and managerial entrenchment purposes.
Indeed, Morck, Shleifer, and Vishny (1990) find poor announcement returns for acquirers who engage in
diversifying acquisitions. Denis, Denis, and Sarin (1997) also find evidence that managerial agency
issues are responsible for firms maintaining value-destroying diversification strategies.
Managers may seek diversification because of the prestige and increased compensation associated
with running bigger firms. Jensen’s (1986) agency theory of free cash flow predicts that when the level
of internal firm resources relative to investment opportunities is higher than expected, managers tend to
overinvest and that the diversification undertaken under such conditions is likely to be value-destroying.
Consistent with this, Lang, Stulz, and Walkling (1991) find that the acquisitions of cash-rich, low-growth
firms perform worse, and Harford (1999) documents that cash-rich firms are more likely to undertake
diversifying acquisitions. However, Stein (2003) stresses the fact that not all firms are empire-building,
and that in only some states of the world would value-destroying overinvestment be present.
Based on
Jensen’s (1986) theory, we do not hypothesize that all diversifications are efficiency-reducing, rather
only the diversifications undertaken by firms with financial slack and with relatively low growth are
likely to have a negative impact on productivity.
2.2.
Regulatory Background
Although the Public Utility Holding Company Act (PUHCA) passed in 1935, gave the SEC the
authority to limit diversification activities by utilities “to such other businesses as are reasonably
incidental, or economically necessary or appropriate” to the operations of the utility, in reality most
utilities were successful in avoiding this law by forming exempt holding companies. The exemption was
-7-
readily granted if the operating utility and non-utility companies of a parent were organized to operate
within the jurisdiction of a single state-level Public Utility Commission (PUC). In fact, the regulation
of diversification has long come under the purview of state-level PUC’s. Further deregulation was
enacted through The Energy Policy Act of 1992 (EPACT), which permitted the formation of unregulated
generation plants and sale at wholesale prices. In general, with a passage of EPACT there was a change
in the mindset regarding the regulation of diversification.
routinely process diversification requests from utilities.
Many states developed procedures to
Regulators were not concerned about the
profitability (or productivity effects) of diversification, but only with protecting ratepayers from having
to cross-subsidize non-utility businesses in their electricity bills.
Since the period of this study,
1990-2003, largely falls after the year 1992, diversification decisions of utilities are not markedly
different from those of other firms.
Nevertheless, in our analysis we control for the ease of state-level
regulation of diversification.
3. Data and Methodology
The data are obtained from a number of sources. Data on the electric utilities are obtained from
Federal Energy Regulation Commission (FERC) Form 1 supplemented with data from POWERdat.
FERC Form 1 is the Annual Report of Major Electric Utilities, downloadable from
http://www.ferc.gov/docs-filing/eforms.asp.
FERC Form 1 contains financial and operational
information filed by the electric utilities themselves.
POWERdat, a comprehensive database for the
electric utilities industry, is provided by Platts, a division of McGraw-Hill Companies, Inc.
POWERdat provides detailed historical information on over 5,000 electric power companies, their
plants and even down to their units. The information provided includes production costs, operation
costs, financial data, investment outlays, electricity sales, fuel supply, rates, wholesale power
transactions, etc.
The level of detail allows us to estimate the productivity of the utilities very
precisely. The sample consists of 1,736 firm-year observations for 124 investor-owned electric power
utilities (electric utility operating companies) in the U.S. during the period from 1990 through 2003.
-8-
The operating companies are required to have total retail sales volume of more than 100,000 Mega Watt
hours (MWh) per year. Companies which are not involved in the retail business, such as those that are
solely in the generation business or wholesale electricity business, are therefore excluded.
We match each operating company to its parent company using information from company
websites and company 10-K filings.
from these sources as well.
Changes in ownership of the operating company are obtained
Sometimes, the parent is acquired and we check through the Mergers and
Acquisitions database of Securities Data Corporation (SDC) to find such changes in ownership.
Accounting information on the parent companies is then obtained from Compustat.
Insert Table 1 around here.
Table 1 provides summary statistics about the operating companies and their corresponding
parents (holding companies) that are included in our sample. The table also provides the states where
the operating companies are located.
Operating companies that are associated with more than one
parent are the ones that changed their parent during the sample period.
Thus the 122 operating
companies are matched to 98 distinct parents, since 2 operating companies cannot be matched to parent
firms on Compustat during our sample period. The two operating companies are Alaska Electric
Light & Power Co. and Mount Carmel Public Utility Co.
The resulting dataset is an unbalanced panel
due to the unavailability of the data for each parent company during each year of our sample period.
3.1.
Measuring Productivity - Total Factor Productivity (TFP)
Our primary measure of firm performance is total factor productivity (TFP) of the operating unit.
The production function is specified by a translog functional form, which essentially is a second-order
approximation to the first-order Cobb-Douglas production function.
Because it incorporates all
second-order (interaction- or cross-) terms across inputs, it is deemed to be very flexible, allowing for
representation of substitution possibilities without restrictive assumptions about the shape of the
-9-
technological relationship.
Let us consider a production of Y with I inputs ( i, j = 1,..., I ). The general formulation of
the translog production function can be mathematically described as follows:
ln Y = α +
∑
I
β i ln xi +
i =1
1
2
∑∑ γ
I
∑∑ γ
I
ij
ij
ln xi ln x j ,
(1.)
i =1 j =1
where the symmetry condition of cross-terms, i.e., γ ij = γ
I
I
ji
for all i and j, and summing up condition,
= 1 for all i, are imposed on parameters.
i =1 j =1
TFP measures are obtained at the operating firm level by estimating the translog production
function for each year (t =1,…,T) across all operating firms (n =1,…, N) in our sample. Specifically,
we estimate the below regression for each year:
1
2
ln Ynt = α t + β vt ln xvnt + β ft ln x fnt + γ vvt (ln xvnt )
2
1
2
+ γ vft ln xvnt ln x fnt + γ fft (ln x fnt ) + ent ,
2
(2.)
t = 1,...,T ,
where Ynt is the output generated by firm n in year t, and xvnt and x fnt are the input factors used in
the production of Ynt .
TFP is then calculated as the estimated residuals from Equation (2)
(Lichtenberg (1992, Ch.2)).
By construction, the estimated residual of operating firm n in year t
measures the percentage deviation of firm n’s TFP from the mean TFP of all firms in year t and can also
be considered as the relative productivity rank of an operating firm in a year. Estimating TFP yearly
has the advantage that the coefficients on the input factors are allowed to vary by year, therefore,
capturing any impact changes in input technology have on productivity.
Due to the detailed nature of the data, we are able to measure the actual quantities of output
that each operating firm generates.
Therefore, output, Ynt is defined as the total volume of electricity
delivered to final customers, measured in Mega Watt hours (MWhs). The two factor inputs are electric
- 10 -
operation and maintenance (O&M) costs and capital stock.
The electric O&M costs include
expenditures for fuel, labor, and purchased power, which are commonly recognized as inputs in the
production of electricity.
The O&M cost is converted to real terms using a state-level gross domestic
product (GDP) deflator, obtained from the Bureau of Economic Analysis (BEA), U.S. Department of
Commerce (downloadable from http://www.bea.gov/regional/gsp/). An advantage of using this deflator
is that we can obtain the time series deflators for each state. However, the disadvantage of using such a
deflator is its generality because the GDP by state includes all industries so that influences of other
industries, that are unrelated to the utilities industry, are included in the movement of the index.
Finally,
capital stock is constructed using the perpetual inventory method. Capital stock for the base year is
calculated by applying a “triangularized” weighted average procedure proposed by Cowing et al. (1981).
The values of capital stock are then written forward each year with the nominal capital expenditures and
depreciation level obtained from the BEA. Such a procedure helps reduce the impact of potential
accounting manipulations of book values of capital stock.
Additional details on the construction of the
capital stock are described in Appendix A.
Table 2, Panel A provides the summary statistics of the calculated TFP and the variables used
in the calculation of TFP.
By construction, the average TFP is zero since TFPs are the estimated
residuals from yearly regressions.
Importantly, there is much variation in the productivity of the
operating firms in our sample as shown by the standard deviation.
Therefore, the efficiency of
operating companies differs greatly, allowing us to have a meaningful test of our hypothesis that parent
firm diversification and financial conditions affect the productivity of the operating unit.
Insert Table 2 around here.
3.2.
Data on Diversification and Financial Constraints
Our main independent variables which measure the degree of diversification of the parent
company are from the Compustat segment files.
We have four measures of diversification. The first
- 11 -
measure is the natural logarithm of number of segments.
To take into account the relative size of the
segments, we calculate herfindahl-based measures of diversification. The second and third measures
are the sales herfindahl and asset herfindahl. Sales herfindahl (asset herfindahl) is the sum of the
squares of each segment sales (assets) as a fraction of total sales (assets). For ease of interpretation, we
use (1-sales herfindahl) and (1-asset herfindahl) in our regressions.
Therefore, the greater the
diversification, the greater will be the values of (1-sales herfindal) and (1-asset herfindahl). The last
measure of diversification is sales from non-utility segments (with SIC codes other than 4911 and 4931)
as a fraction of total sales2.
Jensen (1986) argues that managers are more likely to undertake value-destroying investments if
they have large amounts of free cash flows under their control.
This is especially problematic for low
growth firms which have relatively few positive net present value projects.
By using
internally-generated resources as opposed to raising money in the capital markets, managers can escape
the scrutiny of the capital markets, allowing them to invest in wasteful, unprofitable investments. One
such type of investment is investing in businesses that are unrelated to the core businesses. The
negative value premium for diversified firms has often been attributed to managerial agency problems
(see e.g., Denis, Denis, and Sarin (1997)). Harford (1999) finds that cash-rich firms are more likely to
undertake diversifying acquisitions.
However, Harford does not examine whether the diversifying
acquisitions undertaken by cash-rich firms have differential performance compared with the diversifying
acquisitions undertaken by non-cash-rich firms.
are value-destroying.
It is therefore unclear whether all diversifying actions
We hypothesize that diversification undertaken by firms with large amounts of
free resources relative to investment opportunities will perform worse than diversification undertaken by
managers who are constrained by the amount of resources they have, relative to their growth
opportunities.
To measure whether managers have excess internal cash flows for investments in value-destroying
2
SIC code 4911 is Electric Services industry and SIC code 4931 is Electric and other Services
Combined industry.
- 12 -
projects, we make use of the financial constraints index developed by Kaplan and Zingales (1997),
henceforth called KZ-index. This measure has been adapted for use in several studies such as Lamont,
Polk, and Saa-Requejo (2001), Baker, Stein, and Wurgler (2003), and Malmendier and Tate (2005),
therefore providing external validity for the KZ-index.
Kaplan and Zingales (1997) generate measures
of the degree of financial constraint using quantitative and qualitative information from annual reports to
classify their sample of firms either as financially constrained or financially unconstrained. They then
estimate an ordered logit model on five accounting measures that attempt to capture the degree of
financial constraints: cash flow to net property, plant, and equipment (net PPE), Tobin’s q (Q), leverage
(Leverage), dividends to net PPE (Dividends), and cash holdings to net PPE (CF).
Using the
coefficients from their ordered logit regressions, we construct an index of financial constraints as
follows:
KZ it = −1.001909*
CF it
+ 0.2826389* Qit + 3.139193* Leverageit
PPE it−1
(3.)
Dividendsit
Cashit
− 39.3678*
− 1.314759*
.
PPE it−1
PPE it−1
By construction, higher values of the KZ-index indicates more severe financial constraints than
lower values of the KZ-index.
Based on the KZ-index, financially constrained firms do not have
enough internally-generated cash flows or cash holdings to meet their investments needs as proxied by
their Tobin’s Q. Furthermore, financially constrained firms have high levels of debt, making it difficult
for them to raise funds further from the debt market.
Finally, firms which can afford to pay dividends
are considered not constrained, constrained firms are short on cash and would not pay dividends.
The KZ-index has some attractive features from our perspective.
Based on accounting
information, the KZ-index aggregates various dimensions of being constrained into a single metric, thus
allowing for an objective test of the hypothesis and easier interpretation of the results.
Furthermore,
Jensen’s empire-building hypothesis applies to firms with low growth and high free cash flow. Thus,
the KZ-index already takes into account the growth opportunities of firms together with the amount of
- 13 -
cash flow.
The traditional method of accounting for low growth and high free cash flow is to allow for
an interaction term between low growth and measures of free cash flow (see e.g., Lang, Stulz, and
Walkling (1991)), which can become cumbersome when we also want to take into account other aspects
of being constrained such as leverage ratios, dividend payout, etc.
Furthermore, the KZ-index takes
into account the cash that firms have on hand in addition to the cash flow generated.
Both the amount
of money in the bank and the cash flow generated is likely to affect the amount of resources available to
the manager for investment purposes. Thus the KZ-index is a parsimonious yet comprehensive way to
summarize the financial and investment position of the firm. In untabulated results, we also measure
financial constraints using interest coverage. The results are robust to this alternative measure.
3.3.
Methodology – Estimating the Impact of Diversification on Productivity
This study uses TFP as a comprehensive index representing operating company’s performance.
To examine the relationship between productivity of the operating firm (TFP) and the degree of
diversification (DIV) at the parent firm we estimate the following regression:
TFPnt = a + b1 (OPSize nt ) + b2 (PSize nt ) + b3 (Div nt ) + ε nt ,
(4.)
where OPSize and PSize are size variables to control for differences in the size of the operating company
and parent company, respectively.
OPSIZE (PSize) is the natural logarithm of the sales of the operating
company (parent company) normalized by the average sales of the operating firms (parent company) in
the sample, and
ε nt is an error term.
We take two approaches in estimating the regressions.
In the first approach, we use the
Fama-MacBeth procedure (Fama and MacBeth (1973)) where the coefficients from year-by-year
cross-section regressions are averaged to determine the effects of diversification on productivity, and the
time-series standard errors of the average coefficients are used to draw inferences. In the second
approach, we make use of the fixed effects estimation method, where we include operating firm dummy
variables in the above regression estimations. Two sources of variation in diversification remains after
- 14 -
introducing the operating firm fixed effects: 1) the parent changing its degree of diversification, and 2)
the operating firm moving from a less diversified parent to a more diversified parent or vice versa.
One of the major differences between the two approaches is that the Fama-MacBeth estimation
method makes use of the variation in the degree of diversification across operating companies, ignoring
the variation within the operating firms, while the fixed effects estimation makes use of within-firm
variation, ignoring the variation across firms.
approaches.
It is useful to contrast the results across the two
The Fama-MacBeth results tell us whether diversification affects productivity in the
cross-section, but it is silent on how changes in diversification affects productivity over time. However,
the fixed effects estimation tells us how changes in diversification at the parent firms affect the
productivity of the operating firm, but the fixed effects estimation would not be able to distinguish
whether diversified firms have differential productivity from focused firms at any point in time. This
distinction is important especially in light of evidence from Schoar (2002) that diversified firms are more
productive than focused firms on average, but that they experience a decrease in productivity when they
engage in diversification activities.
This difference in the cross-section and time series would be lost if
we simply estimate a pooled regression that takes into account both the within- and between-firm effects.
Table 2, Panel B provides some summary statistics of the independent variables used in this study.
Consistent with the deregulation efforts during the 1980s and early 1990s, there are varying degrees of
diversification among the parent companies; the minimum number of segments is one, with the
maximum being nine segments.
The median parent company has three segments. The other three
measures of diversification also show that there is variation in the degree of diversification among the
parent companies.3
The average (median) percentage of sales from non-utility segments is about 16%
(9%). Thus most of the parent companies are mainly involved in the utility business. Our measure of
financial constraints, KZ-index, also shows that there is variation in whether parent firms are financially
constrained or not.
3
The minimum non-utility sales is negative because of negative net sales in a segment (SIC code =
4813) belonging to parent company, Northeast Utilities.
- 15 -
4. Empirical Results
4.1.
Productivity and Diversification
Table 3 examines the impact of diversification on productivity using the specification in Equation
(4).
Panel A shows cross-sectional effects of diversification on TFP, where the regressions are
estimated using Fama-Macbeth regressions.
These estimations examine how the diversification
measures affect firms’ TFP at a given point in time, only making use of the cross-sectional variation in
the data. Model 1 uses the logarithm of number of segments as a diversification measure, Model 2 uses
(1 - sales herfindahl), Model 3 uses (1 - assets herfindahl), and Model 4 uses the ratio of revenue from
non-utility businesses to the total revenue of the parent company. All models show that diversified
parents have less productive utility businesses. The coefficients on the diversification measures are all
negative and significant at the 1% level. Also, the results are economically significant, a one standard
deviation increase in (1-sale herfindahl) leads to a 3% drop in productivity. This result is consistent
with the literature that there exists a diversification discount. Using plant-level data from the U.S.
Census Bureau, Maksimovic and Phillips (2002) also find that conglomerate firms are less productive
than single-segment firms.
Insert Table 3 around here.
However, it is unclear from such cross-sectional regressions whether the negative coefficients on
the diversification variables are due to the fact that diversification reduces productivity or whether
parents with less productive operating companies choose to diversify.
Papers such as Campa and Kedia
(2002) and Villalonga (2004) argue that the conglomerate discount arises endogenously. Consistent
with this argument, Jandik and Makhija (2005) find that underperforming electric utilities are more
likely to diversify.
The model in Maksimovic and Phillips (2002) shows that the reduced productivity
of conglomerate firms relative to single-segment firms can be consistent with profit-maximization.
- 16 -
In
their model, firms have differential comparative advantage across different industries. Therefore, firms
which are relatively more productive in a specific industry have higher opportunity costs of diversifying
and thus, in equilibrium, single-segment firms have higher productivity than conglomerates. In their
model, it may be optimal for a firm which is relatively less efficient in a specific industry to diversify
into other industries.
Thus, it is possible that the time-series dynamic relation between diversification
and productivity may be different compared to the cross-sectional relation. Therefore, we next examine
whether diversifying activities lead to a worsening of productivity at the operating company or not.
We estimate Equation (4) using fixed effects estimation in Panel B. By including operating firm
fixed effects, not only can we examine the temporal effects of diversification on productivity, we can
also control for any unobserved characteristics of the operating company that can affect the relation
between productivity and diversification.
In contrast to the cross-sectional results in Panel A, the
results in Panel B indicate that increases in parent’s degree of diversification positively influence the
TFP of their operating companies.
Taken together, Table 3 shows that diversified parent companies have lower productivity in their
operating companies as compared to single-segment parent companies. However, diversification into
other non-utility-related segments in fact helps to increase the productivity of their utility segment.
Our results so far are different from the results in Schoar (2002), who finds that diversified firms
have higher productivity in the cross-section, but the act of diversification reduces productivity. Our
results are however consistent with Maksimovic and Phillips (2002) who find that diversified firms are
less productive than single-segment firms of similar size, except for the smallest firms. One reason for
the difference in results between Schoar (2002) and ours could be due to the differing conditions under
which firms in her sample and our sample undertake their diversifying activities. Jensen (1986) argues
that investments undertaken by cash-rich, low-growth firms are likely to be value-destroying.
Therefore, diversification activities may not always be efficiency-decreasing, the productivity effects
depend on whether the firms have excess resources for investments and whether there are good
investment opportunities available.
Therefore, in subsequent sections, we examine whether
- 17 -
diversification undertaken when firms are financially constrained have a different impact on productivity
compared to when firms are not financially constrained. The impact of financial constraints on the
productivity and value effects of diversification has not been examined before.
4.2.
Productivity, Diversification, and Financial Constraints
We first examine the impact of financial constraints on productivity.
The specification we
estimate is:
TFPnt = a + b1 (OPSizent ) + b2 ( PSizent ) + b3 (Divnt ) + b4 ( KZ nt −1 ) + ε nt ,
(5.)
where KZ nt −1 is the KZ-index from Kaplan and Zingales (1997) and measures the degree of financial
constraints and financial slack the parent company faces. Higher values of the KZ-index indicate
higher financial constraints and lower slack.
Insert Table 4 around here.
Table 4 summarizes estimation results of Equation. (5).
Panel A provides cross-sectional results
using Fama-MacBeth regressions, while Panel B shows the dynamic effects of financial constraints on
TFP using fixed effects estimations.
The results show that financial constraints negatively affect TFP
both in the cross-section and in the time series. Companies faced with financial constraints likely do
not have enough resources to invest in profitable projects or new technology therefore resulting in lower
efficiency of the operating firms.
In untabulated results, we measure financial constraints using interest
coverage and the results are generally robust to this alternative measure of financial constraints.
Controlling for the KZ-index in the regressions reduces the statistical significance of the
coefficients on the diversification variables.
Importantly, the signs of the coefficients on the
diversification variables are similar to those in Table 3. In Panel A, diversification is again negatively
associated with productivity in the cross-section, although only the coefficient estimates on the logarithm
of number of segments and (1-asset herfindahl) is significant at conventional significance level.
- 18 -
In
Panel B, the coefficient estimates on diversification are positive for all models. However, only the ratio
of non-utility sales to total sales is significant.
Table 4 does not allow us to determine the impact of financial constraints and financial slack on
the productivity effects of diversification.
To examine whether diversification undertaken when the
firm is financially constrained has a different impact on productivity compared to diversification
undertaken when there is financial slack, we introduce interaction terms between the level of
diversification and the KZ-index. Specifically, we estimate the below specification:
TFPnt = a + b1 (OPSize nt ) + b2 (PSize nt ) + b3 (Div nt ) + b5 (KZH nt −1 ) + b6 (KZLnt −1 )
+ b7 (Div nt )(KZH nt −1 ) + b8 (Div nt )(KZLnt −1 ) + ε nt ,
(6.)
where KZH takes the value of one when the firm’s KZ-index belongs to the top tercile of KZ-index
values in the year, and zero otherwise. In contrast, KZL takes the value of one when the firm’s
KZ-index belongs to the bottom tercile of KZ-index values in the year, and zero otherwise. Since we
are interested in the dynamic effects of diversification on productivity, we estimate Equation (6) using
the fixed effects estimation.
In this study, we refer to firm-years where KZ-index is in the top tercile (KZH equals one) as
being financially constrained while firm-years where the KZ-index is in the bottom tercile (KZL equals
one) are considered as being financially unconstrained.
It is important to note that we do not mean that
firms in the top tercile are completely constrained while firms in the bottom tercile are completely
unconstrained. This segregation of the firm-years based on their KZ-index merely implies that firms in
the top tercile are relatively more constrained than firms in the bottom tercile.
Therefore, b7
measures the impact of diversification on productivity for relatively more constrained firms while b8
measures the impact of diversification on productivity for firms with relatively more financial slack.
Our hypothesis predicts that b7 will be greater than b8 , i.e., diversification undertaken during
financial slack has a negative impact on productivity compared to diversification undertaken when the
firm is relatively more constrained.
- 19 -
Insert Table 5 around here.
Table 5 summarizes the estimation results of Equation (6) obtained from fixed effects estimation.
First, all the diversification measures are positive and significant for all models. This is consistent with
the results in Panel B of Table 3. Second, the coefficient estimates of KZH are negative and those of
KZL are positive in all the models. The coefficients on KZH and KZL are all statistically significant at
conventional levels with the exception of KZH of Model 1. Consistent with results in Table 4, the
results show that severe financial constraints are associated with decreases in TFP, while moderate
financial constraints or financial slack is associated with increases in TFP.
More interestingly, we find
that the negative impact of diversification on productivity is only among firms with financial slack.
The interaction term between KZL and the diversification variables are all negative and significant for
the first three models.
There is no such negative impact of diversification on productivity for the
financially constrained firms.
In fact, the coefficients on the interaction term between the
diversification variables and KZH are generally positive, although insignificant.
We test for whether
the coefficient estimates on the interaction terms (i.e., b7 and b8 ) are significantly different from each
other.
The F-statistics and the corresponding P-values are listed at the bottom of Table 5. All the
F-statistics reject the null hypothesis that the coefficient estimates are the same between cross-terms of
the diversification measure with KZH and that with KZL, further supporting the hypothesis that different
financial conditions have an impact on whether diversification activities are good or bad for the firm.
In untabulated results, we replicated Table 5 using interest coverage as an alternative measure of
financial constraint and the results are similar.
These results also reconcile Table 3 with results in Schoar (2002) who finds that diversification
activities lead to a negative impact on the productivity of incumbent plants. She argues that this is
because of the “new toy” effect.
Managers who diversify shift their focus to the new segment,
therefore neglecting the incumbent segments.
This neglect by management leads to reduction in the
- 20 -
productivity of the incumbent segments.
In Table 5, we find that diversification undertaken when firms
have excess resources relative to investment opportunities leads to negative impact on the productivity
of the incumbent utility segment, but when firms are financially constrained, diversification in fact has a
positive effect on the productivity of the utility segment. Therefore, when considering the impact of
diversification on productivity, it is important to consider the financial conditions under which the
diversification activities are taken.
4.3.
Robustness check
Fixed effects estimation makes use of the within-firm variation to derive the coefficient estimates,
thus including operating companies which do not experience any changes in diversification would
introduce noise into the estimation.
As can be seen from Table 6, deleting operating companies which
do not experience any changes in diversification leads to stronger results. The absolute value of the
coefficients on the interaction term between diversification and KZL are slightly larger than those in
Table 5 and now the interaction term between the ratio of non-utility sales to total sales is significant at
the 10% level.
Insert Table 6 around here.
Even though we study a period during which diversification activities are not seriously regulated,
we allow for the possibility that there are other regulatory effects. In Table 7, we examine whether
deregulation efforts in the utility industry would affect our results. The year, 1996, is important from
the perspective of industrial policy for the electric utilities, because FERC issued Orders 888 and 889
that allowed market participants for accessing transmission lines to promote wholesale power trading
and foster competition in the industry. In addition, restructuring legislations on the electric utility
industry were passed in some states (e.g., Rhode Island and California) in 1996. Hence, we introduce
additional variables to control for effects of deregulation. Reg96 is a dummy variable that takes the
- 21 -
value of one during the period from 1996 to 2003, and zero otherwise.
diversification variables.
We also interact Reg96 with the
Again, the conclusion remains the same even after controlling for
deregulation. In particular, the deregulation variable and its cross-term with the diversification measure
are not statistically significant.
Insert Table 7 around here.
5. Conclusion
The electric utility industry provides for a natural setting to test for the impact of diversification
on productivity.
Electric utilities produce a single homogenous product, which enables precise
measurement of productivity and comparison of productivity across firms.
Also, the industry is large
enough with varying degrees of diversification across firms to viably test for the impact of
diversification on productivity.
Using data from the period of deregulation, we find that operating
companies of diversified parent firms have lower productivity than those of focused parent firms in the
cross-section.
However, in the time series, we find that productivity in the core utility segment
increases when the parent company diversifies.
This increase in productivity is only among firms with
relatively high and moderate levels of financial constraints.
When parent firms with financial slack
diversify, we find that the productivity of the operating company falls.
This suggests that when
managers are faced with excess resources relative to their investment opportunities, they tend to invest in
“pet projects” which distracts them from their core business, leading to reductions in the productivity of
the electric utility operating company.
- 22 -
Appendix A: Construction of Capital Stock Data
This study constructs the capital stock data employing a perpetual inventory method. Capital
stock in current year t is defined as such:
CS it = CS it −1 (1 − δ ) +
GI it
,
PI t
(A.1)
where CS it is the real-term capital stock for firm i in period t, GI it is the nominal-term gross capital
investments, and PI t is a price index for converting nominal terms to real terms. GI it is calculated
by summing up all gross additions to utility capital assets, obtained from the operating firm’s cash flow
statement.
The depreciation rate, δ , is from the Bureau of Economic Analysis (BEA) at the U.S.
Department of Commerce, downloadable from http://bea.gov/bea/an/0597niw/tableA.htm. Specifically,
we use a constant depreciation rate of 0.0211 over the sample period, which is constructed for the
－electric light and power.”
category of “private nonresidential structure
The capital stock for the base-year, i.e., 1990, is constructed by applying a “triangularized”
weighted average procedure proposed by Cowing et al. (1981):
CSib =
∑
BK ib
⎧⎛
⎨⎜ r
r =1 ⎩⎝
20
∑
⎞ ⎫
r ⎟ PI r ⎬
r =1 ⎠
⎭
20
, b = 1990,
(A.2)
where BKib is the book value of capital assets for firm i in the base-year. PI 1 to PI 20 corresponds to
PI 1971 to PI 1990 , respectively.
The price index, PI, used in Equations A.1 and A.2 is obtained from the Price Indexes for
Gross Domestic Product Table published by the BEA. Specifically, we apply the decomposed index
constructed for the category of “Gross private domestic investment -- Fixed investment -- Nonresidential
structures.”
This index is downloadable from
http://www.bea.gov/national/nipaweb/TableView.asp?SelectedTable=4&FirstYear=2004&LastYear=
2006&Freq=Qtr.
- 23 -
References
Baker, Malcom, Jeremy C. Stein, and Jeffrey Wurgler, 2003, When Does the Market Matter? Stock
Prices and the Investment of Equity-Dependent Firms, The Quarterly Journal of Economics
118, 969-1005.
Berger, Philip G., and Eli Ofek, 1995, Diversification’s Effect on Firm Value, Journal of Financial
Economics 37, 39-65.
Campa, Jose Manuel, and Simi Kedia, 2002, Explaining the Diversification Discount, Journal of Finance
57, 1731-1762.
Comment, Robert, and Gregg A. Jarrell, 1995, Corporate Focus and Stock Returns, Journal of Financial
Economics 37, 67-87.
Denis, David J., Diane K. Denis, and Atulya Sarin, 1997, Agency Problems, Equity Ownership, and
Corporate Diversification, Journal of Finance 52, 135-160.
Fama, Eugene F., and James D. MacBeth, 1973, Risk, Return, and Equilibrium: Empirical Tests, Journal
of Political Economy 81, 607-636.
Graham, John R., Michael L. Lemmon, and Jack G. Wolf, 2002, Does Corporate Diversification Destroy
Value?, Journal of Finance 57, 695-720.
Harford, Jarrad, 1999, Corporate Cash Reserves and Acquisitions, Journal of Finance 54, 1969-97.
Jandik, Tomas, and Anil K. Makhija, 2005, Can Diversification Create Value? Evidence from the
Electric Utility Industry, Financial Management 34, 61-93.
Jensen, Michael C., 1986, Agency Costs of Free Cash Flow, Corporate Finance and Takeovers, American
Economic Review 76, 323-329.
Jensen, Michael C., 1993, The Modern Industrial Revolution, Exit, and the Failure of Internal Control
Systems, Journal of Finance 48, 831-880.
Jensen, Michael C., and William H. Meckling, 1976, Theory of the Firm: Managerial Behavior, Agency
Costs, and Ownership Structure, Journal of Financial Economics 3, 305-360.
- 24 -
Joskow, P. (2000), “Deregulation and Regulatory Reform in the US Electric Power Sector,” in
Deregulation of Network Industries: What’s Next? ed. Sam Peltzman and Clifford Winston,
113-188. Washington, D.C.: AEI-Brookings Joint Center for Regulatory Studies, Brookings
Institution Press.
Kaplan, Steven N., and Luigi Zingales, 1997, Do Investment-Cash Flow Sensitivities Provide Useful
Measures of Financing Constraints?, Quarterly Journal of Economics 112, 169-215.
Lamont, Owen, Christopher Polk, and Jesus Saa-Requejo, 2001, Financial Constraints and Stock Returns,
Review of Financial Studies 14, 529-554.
Lang, Larry H. P., and René M. Stulz, 1994, Tobin’s Q, Corporate Diversification and Firm Performance,
Journal of Political Economy 102, 1248-1280.
Lang, Larry H. P., Rene M. Stulz, and Ralph A. Walkling, 1991, A Test of the Free Cash Flow
Hypothesis: The Case of Bidder Returns, Journal of Financial Economics 29, 315-35.
Lins, K. V., and H. Servaes, 2002, Is Corporate Diversification Beneficial in Emerging Markets?,
Financial Management 31, 5-31.
Lins, Karl, and Henri Servaes, 1999, International Evidence on the Value of Corporate Diversification,
Journal of Finance 54, 2215-2239.
Maksimovic, Vojislav, and Gordon Phillips, 2001, The Market for Corporate Assets: Who Engages in
Mergers and Asset Sales and Are There Efficiency Gains? Journal of Finance 56, 2019-65.
Maksimovic, Vojislav, and Gordon Phillips, 2002, Do Conglomerate Firms Allocate Resources
Inefficiently across Industries? Theory and Evidence, Journal of Finance 57, 721-767.
Malmendier, Ulrike M., and Geoffrey Tate, 2005, CEO Overconfidence and Corporate Investment,
Journal of Finance 60, 2661-2700.
Morck, Randall, Andrei Shleifer, and Robert W. Vishny, 1990, Do Managerial Objectives Drive Bad
Acquisitions?, Journal of Finance 45, 31-48.
Schoar, Antoinette, 2002, Effects of Corporate Diversification on Productivity, Journal of Finance 57,
2379-2403.
- 25 -
Servaes, Henri, 1996, The Value of Diversification During the Conglomerate Merger Wave, Journal of
Finance 51, 1201-1225.
Sing, M. (1987), “Are Combination Gas and Electric Utilities Multiproduct Natural Monopolies?”
Review of Economics and Statistics, 69 (3), 392-398.
Stein, Jeremy C., 2003, Agency, Information and Corporate Investment, Handbook of the economics of
finance. Volume 1A. Corporate finance (Elsevier, North Holland, Amsterdam; London and
New York).
Villalonga, Belen, 2004, Does Diversification Cause the 'Diversification Discount'?, Financial
Management 33, 5-27.
- 26 -
Table 1: List of sample firms
The table lists the investor-owned electric power utilities (IOUs) and their parent firms in our sample. IOUs corresponding to more than
one parent changed ownership during the sample period. Parent firms are identified using information from company websites and 10-K
filings. Changes in ownership of the IOUs are also obtained from SDC’s Mergers and Acquisitions database.
Number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
Company Name
AEP Texas Central Co.
Parent Company
STATE Number
Central and South West Corporation OH
American Electric Power Co., Inc.
AEP Texas North Co.
Central and South West Corporation OH
American Electric Power Co., Inc.
Alabama Power Co.
Southern Co.
AL
Alaska Electric Light & Power Co.
AK
Appalachian Power Co.
American Electric Power Co., Inc.
OH
Aquila Inc.
Aquila, Inc
MO
Arizona Public Service Co.
Pinnacle West Capital Corp.
AZ
Atlantic City Electric Co.
Atlantic Energy Inc
NJ
Conectiv
PEPCO Holdings, Inc.
Avista Corp.
Avista Corp.
WA
Baltimore Gas & Electric Co.
Constellation Energy Group, Inc.
MD
Bangor Hydro-Electric Co.
BANGOR HYDRO ELECTRIC CO ME
Black Hills Power Inc.
Black Hills Corp.
SD
Boston Edison Co.
NSTAR
MA
Cambridge Electric Light Co.
Commonwealth Energy System
MA
NSTAR
Carolina Power & Light Co.
Progress Energy, Inc.
NC
CenterPoint Energy Houston Electric, LLCCenterPoint Energy Inc.
TX
Central Hudson Gas & Electric Corp.
CH Energy Group, Inc.
NY
Central Illinois Light Co.
CILCORP Inc
IL
AES Corp.
Ameren Corp.
Central Illinois Public Services Co.
C I P S C O INC
IL
Ameren Corp.
Central Maine Power Co.
C M P GROUP INC
ME
Energy East Corp.
Central Vermont Public Service Corp.
Central Vermont Public Service Corp VT
Cincinnati Gas & Electric Co.
Cinergy Corp.
OH
Clark Fork & Blackfoot, LLC
Montana Power Co
MT
NorthWestern Corp.
Cleco Power LLC
Cleco Corp.
LA
Cleveland Electric Illuminating Co.
Centerior Energy
OH
FirstEnergy Corp.
Columbus Southern Power Co.
American Electric Power Co., Inc.
OH
Commonwealth Edison Co.
UNICOM CORP HOLDING CO
IL
Exelon Corp.
Commonwealth Electric Co.
Commonwealth Energy System
MA
NSTAR
Connecticut Light & Power Co.
Northeast Utilities
CT
Connecticut Valley Electric Co., Inc.
Central Vermont Public Service Corp CT
Consolidated Edison Co. Of New York IncConsolidated Edison, Inc.
NY
Consumers Energy Co.
CMS Energy Corp.
MI
Dayton Power & Light Co.
DPL, Inc.
OH
Delmarva Power & Light Co.
Conectiv
DE
PEPCO Holdings, Inc.
Detroit Edison Co.
DTE Energy Co.
MI
Duke Energy Indiana, Inc.
PSI Resources Inc
IN
Cinergy Corp.
Duke Power Co.
Duke Energy Corp.
NC
Duquesne Light Co.
Duquesne Light Holdings, Inc.
PA
Company Name
39
Edison Sault Electric Co.
40
41
42
43
44
45
46
47
48
El Paso Electric Co.
Electric Energy, Inc.
Empire District Electric Co.
Entergy Arkansas, Inc.
Entergy Gulf States, Inc.
Entergy Louisiana, Inc.
Entergy Mississippi, Inc.
Entergy New Orleans, Inc.
Fitchburg Gas & Electric Light Co.
49
50
Florida Power & Light Co.
Florida Power Corp.
Parent Company
ESELCO Inc
Wisconsin Energy Corp.
El Paso Electric Co.
Ameren Corp.
Empire District Electric Co.
Entergy Corp.
Entergy Corp.
Entergy Corp.
Entergy Corp.
Entergy Corp.
Fitchburg Gas & Electric Lt.Co.
Unitil Corp.
FPL Group, Inc.
Florida Progress Corporation
Progress Energy, Inc.
Southern Co.
Green Mountain Power Corp.
Southern Co.
Hawaiian Electric Industries, Inc.
IDACORP, Inc.
ILLINOVA CORP
Dynergy Inc
American Electric Power Co., Inc.
IPALCO Enterprises Inc.
AES Corp.
Great Plains Energy Corp.
American Electric Power Co., Inc.
KU Energy
LG&E Energy
Kansas Gas & Electric Co.
Westar Energy Inc.
American Electric Power Co., Inc.
LG&E Energy
MGE Energy Inc.
Hawaiian Electric Industries, Inc.
ALLETE
Southern Co.
Allegheny Energy, Inc.
MDU Resources Group, Inc.
STATE Number
51
52
53
54
55
56
Georgia Power Co.
Green Mountain Power Corp.
Gulf Power Co.
Hawaiian Electric Co., Inc.
Idaho Power Co.
Illinois Power Co.
57
58
Indiana Michigan Power Co.
Indianapolis Power & Light Co.
59
60
61
Kansas City Power & Light Co.
Kentucky Power Co.
Kentucky Utilities Co.
62
KGE, A Westar Energy Co.
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
Kingsport Power Co.
Louisville Gas & Electric Co.
Madison Gas & Electric Co.
Maui Electric Co., Ltd.
Minnesota Power, Inc.
Mississippi Power Co.
Monongahela Power Co.
Montana Dakota Utilities Co.
Mount Carmel Public Utility Co.
Nevada Power Co.
Sierra Pacific Resources
New York State Electric & Gas Corp. Energy East Corp.
Northern Indiana Public Service Co. NiSource Inc
Northern States Power Co.
Xcel Energy, Inc.
Northern States Power Co. Wisconsin Xcel Energy, Inc.
NorthWestern Energy,
NorthWestern Corp.
a Division of Northwestern Co
Ohio Edison Co.
FirstEnergy Corp.
Ohio Power Co.
American Electric Power Co., Inc.
Oklahoma Gas & Electric Co. (OG&EOGE Energy Corp.
Orange & Rockland Utilities, Inc.
ORANGE & ROCKLAND UTILS INC
Consolidated Edison, Inc.
Otter Tail Power Co.
Otter Tail Corp.
Pacific Gas and Electric Co.
PG&E Corp
82
83
- 27 -
MI
TX
IL
MO
AR
TX
LA
MS
LA
MA
FL
FL
GA
VT
FL
HI
ID
IL
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
OH
IN
MO
OH
KY
100
101
102
103
104
105
KS
106
TN
KY
WI
HI
MN
MS
WV
ND
IL
NV
NY
IN
MN
WI
SD
OH
OH
OK
NY
ND
CA
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
Company Name
PacifiCorp
PECO Energy Co.
Pennsylvania Electric Co.
Parent Company
STATE
PACIFICORP
OR
Exelon Corp.
PA
GPU Inc
OH
FirstEnergy Corp.
Pennsylvania Power Co.
FirstEnergy Corp.
OH
Potomac Edison Co.
Allegheny Energy, Inc.
PA
Potomac Electric Power Co.
PEPCO Holdings, Inc.
DC
PPL Electric Utilities Corp.
PPL Corp.
PA
PSC of Colorado
NEW CENTURY ENERGIES INC
CO
Xcel Energy, Inc.
PSC of New Hampshire
PUBLIC SERVICE CO OF NH
NH
Northeast Utilities
PSC of Oklahoma
Central and South West Corporation
OH
American Electric Power Co., Inc
Public Service Co. of New Mexico PNM Resources
NM
Public Service Electric and Gas Co. Public Service Enterprise Group, Inc.
NJ
Puget Sound Energy, Inc.
Puget Energy, Inc.
WA
Rochester Gas & Electric Corp.
R G S ENERGY GROUP INC
NY
Energy East Corp.
Rockland Electric Co.
ORANGE & ROCKLAND UTILS INC NY
Consolidated Edison, Inc.
San Diego Gas & Electric Co.
ENOVA CORP
CA
Sempra Energy
Savannah Electric & Power Co.
Southern Co.
GA
Sierra Pacific Power Co.
NV
Sierra Pacific Resources
South Beloit Water, Gas & Electric CoAlliant Energy Corp.
WI
South Carolina Electric & Gas Co. SCANA Corp.
SC
Southern California Edison Co.
Edison International
CA
Southern Indiana Gas & Electric Co. SIGCORP Inc
IN
Vectren Corp.
Southwestern Electric Power Co.
Central and South West Corporation
OH
American Electric Power Co., Inc
SOUTHWESTERN PUBLIC SERVICE TX
Southwestern Public Service Co.
NEW CENTURY ENERGIES INC
Xcel Energy, Inc.
Superior Water, Light & Power Co. ALLETE
WI
Tampa Electric Co.
TECO Energy, Inc.
FL
Texas-New Mexico Power Co.
TNP ENTERPRISES INC
TX
Toledo Edison Co.
Centerior Energy
OH
FirstEnergy Corp.
Tucson Electric Power Co
UniSource Energy Corp.
AZ
Union Electric Co.
Ameren Corp.
MO
Union Light, Heat & Power Co.
Cinergy Corp.
KY
United Illuminating Co.
UIL Holdings Corp.
CT
Upper Peninsula Power Co.
UPPER PENINSULA ENERGY CORP MI
WPS Resources Corp.
Virginia Electric & Power Co.
Dominion Resources, Inc.
VA
West Penn Power Co.
Allegheny Energy, Inc.
PA
Westar Energy
Westar Energy Inc.
KS
Western Massachusetts Electric Co. Northeast Utilities
MA
Wheeling Power Co.
American Electric Power Co., Inc.
OH
Wisconsin Electric Power Co.
Wisconsin Energy Corp.
WI
Wisconsin Power & Light Co.
Alliant Energy Corp.
WI
Wisconsin Public Service Corp.
WPS Resources Corp.
WI
Table 2: Descriptive statistics
The table provides the average, median, maximum, minimum, and standard deviation of the variables
used. Panel A provides statistics of the variables used to estimate total factor productivity (TFP). Output
is measured using total volume of electricity sold to final customers measured in Mega Watt hours. The
two input variables are electric operation and maintenance (O&M) cost and capital stock. TFP is defined
by the residuals from estimating Equation (2). The residuals indicate a percentage deviation of that
firm’s TFP from the mean TFP of all firms in total samples. By this definition, the average TFP is close
to 0. Panel B gives the statistics of variables describing the IOU and the parent company. OPSIZE
(PSIZE) is the 3-year moving average of total sales of the IOU (parent company) in million dollars. All
dollar values are in 2002 dollars. Segment data and accounting data for the parent company are from
Compustat. The degree of diversification is measured using NSEG (number of segments), 1-SALEH
(1-sales herfindahl index), 1–ASSETSH (1–assets herfindahl index), and NONUTIL (Sales from
non-utility segments as a fraction of total firm sales). Financial constraints is measured using the Kaplan
and Zingales (KZ) index, defined as
KZ = −1.002*
CF
Dividends
Cash
+ 0.283* Q + 3.139* Leverage − 39.368*
− 1.315*
PPE
PPE
PPE
Panel A. Data for estimating TFP
Electricity Sales
(1,000 MWh)
O&M Cost
(Mio. $)
Capital Stock
(Mio. $)
TFP (%)
Avg.
21,014
876.838
48.917
-1.217E-09
Med.
Max.
Min.
13,916
182,194
143
533.337
8,674.834
2.955
28.174
417.972
0.080
0.029
3.088
-2.867
S.D.
22,163
997.329
61.239
0.499
Panel B. Data for size, diversification, and financial constraints
OPSIZE
PSIZE
NSEG
1-SALEH 1-ASSETSH NONUTIL
KZ
Avg.
1,354
3,838
2.653
0.268
0.262
0.161
0.227
Med.
822
2,425
3.000
0.287
0.241
0.092
0.136
Max.
8,425
41,330
9.000
0.789
0.820
0.962
4.041
Min.
9
29
1.000
0.000
0.000
-0.005
-3.478
S.D.
1,550
4,814
1.534
0.238
0.238
0.200
0.775
- 28 -
Table 3: Effects of diversification on productivity
The table examines the effects of diversification on productivity. The dependent variable is TFP,
defined as the residual from estimating equation (2). In the regressions, OPSIZE is defined as the
natural logarithm of the total sales of the operating company normalized by the average OPSIZE in
the sample. PSIZE is defined as the natural logarithm of the total sales of the parent company
normalized by the average PSIZE in the sample. The degree of diversification is measured using
NSEG (the natural logarithm of number of segments), 1-SALEH (1-sales herfindahl index),
1–ASSETSH (1–assets herfindahl index), and NONUTIL (Sales from non-utility segments as a
fraction of total firm sales). In Panel A, the Fama-MacBeth coefficient estimates are the time-series
average of coefficients from yearly cross-sectional regressions over the sample period. In Panel B,
the fixed effects coefficient estimates are from panel regressions which include operating firm fixed
effects. Absolute t-statistics are given below the coefficient estimates. *, **, and *** indicate
significance at the 10%, 5%, and 1% levels, respectively.
Panel A. Fama-Macbeth regression
Variables
OPSIZE
PSIZE
NSEG
Model 1
Model 2
Model 3
Model 4
Coefficients
t-ratio
0.044
12.39
0.035
5.07
-0.089
-3.19
Coefficients
t-ratio
0.046
10.28
0.030
4.59
Coefficients
t-ratio
0.044
9.58
0.029
4.30
Coefficients
t-ratio
0.041
10.52
0.024
6.87
***
***
***
***
***
***
-0.137
-4.33
***
-0.186
-5.50
1-ASSETSH
***
NONUTIL
R-squared
No. of Obs.
0.110
3.31
0.042
1542
***
***
1-SALEH
CONS
***
***
0.053
5.80
0.040
1540
- 29 -
***
0.058
6.08
0.044
1519
***
-0.288
-8.21
0.049
6.54
0.050
1446
***
***
Panel B. Fixed effects estimation
Variables
OPSIZE
PSIZE
NSEG
Model 1
Model 2
Model 3
Model 4
Coefficients
t-ratio
0.053
1.78
-0.031
3.86
0.027
3.29
Coefficients
t-ratio
0.060
2.04
-0.029
3.77
Coefficients
t-ratio
0.051
1.69
-0.034
4.21
Coefficients
t-ratio
0.067
2.26
-0.036
-4.85
*
***
***
NONUTIL
0.914
1542
***
**
***
***
0.087
3.82
1-ASSETSH
Adj. R-squared
No. of Obs.
*
***
0.077
3.39
1-SALEH
**
0.914
1540
- 30 -
0.915
1519
***
0.055
1.74
0.933
1446
*
Table 4: Effects of diversification and financial constraints on productivity
The table examines the effects of diversification and financial constraints on productivity. The
dependent variable is TFP, defined as the residual from estimating equation (2). In the regressions,
OPSIZE is defined as the natural logarithm of the total sales of the operating company normalized
by the average OPSIZE in the sample. PSIZE is defined as the natural logarithm of the total sales of
the parent company normalized by the average PSIZE in the sample. The degree of diversification
is measured using NSEG (the natural logarithm of number of segments), 1-SALEH (1-sales
herfindahl index), 1–ASSETSH (1–assets herfindahl index), and NONUTIL (Sales from non-utility
segments as a fraction of total firm sales). Financial constraints is measured using the Kaplan and
Zingales (KZ) index. In Panel A, the Fama-MacBeth coefficient estimates are the time-series
average of coefficients from yearly cross-sectional regressions over the sample period. In Panel B,
the fixed effects coefficient estimates are from panel regressions which include operating firm fixed
effects. Absolute t-statistics are given below the coefficient estimates. *, **, and *** indicate
significance at the 10%, 5%, and 1% levels, respectively.
Panel A. Fama-Macbeth regression
Variables
OPSIZE
PSIZE
Model 1
Model 2
Model 3
Model 4
Model 5
Coefficients
t-ratio
0.026
5.33
0.037
4.90
Coefficients
t-ratio
0.030
5.62
0.030
4.07
-0.068
-3.02
Coefficients
t-ratio
0.031
5.56
0.024
2.97
Coefficients
t-ratio
0.030
5.52
0.027
3.93
Coefficients
t-ratio
0.032
5.25
0.021
4.74
***
***
NSEG
***
***
***
**
***
***
-0.031
-0.66
-0.107
-2.30
1-ASSETSH
**
NONUTIL
CONS
R-squared
No. of Obs.
***
***
1-SALEH
KZ
***
-0.080 ***
-4.86
0.040 ***
4.94
0.052
1513
-0.093
-5.57
0.103
3.71
0.052
1434
***
***
- 31 -
-0.085
-5.10
0.037
2.12
0.053
1432
***
*
-0.093
-5.72
0.057
4.56
0.056
1411
***
***
-0.036
-0.49
-0.094
-6.02
0.033
1.78
0.060
1343
***
*
Panel B. Fixed effects estimation
Model 1
Variables
OPSIZE
PSIZE
NSEG
Model 2
Model 3
Model 4
Model 5
Coefficients
Coefficients
t-ratio
t-ratio
***
0.085
0.097
3.18
2.90
-0.001
0.003
-0.13
0.36
0.012
1.61
Coefficients
t-ratio
0.086
2.96
0.004
0.46
Coefficients
t-ratio
0.090
2.98
0.007
0.83
Coefficients
t-ratio
0.098
3.07
-0.021
-2.66
***
***
***
0.016
0.72
1-ASSETSH
NONUTIL
Adj. R-squared
No. of Obs.
***
0.035
1.61
1-SALEH
KZ
***
-0.009
-1.31
0.885
1513
-0.016
-2.54
0.902
1434
**
- 32 -
-0.016
-2.53
0.902
1432
**
-0.021
-3.08
0.902
1411
***
0.065
1.96
-0.024
-3.71
0.916
1343
**
***
Table 5: Impact of financial constraints on relation between diversification and
productivity
The table examines whether financial constraints affect the relation between diversification and
productivity using the fixed effects estimation. The dependent variable is TFP, defined as the
residual from estimating equation (2). KZH (KZL) is a dummy variable that takes 1 when the firm’s
KZ index belongs to the top (bottom) 1/3 of the variable for each year and 0 otherwise. The other
variables are as defined in the legend of Table 2. All models include operating firm fixed effects.
Absolute t-statistics are given below the coefficient estimates. *, **, and *** indicate significance at
the 10%, 5%, and 1% levels, respectively.
Variables
OPSIZE
PSIZE
NSEG
Model 1
Model 2
Model 3
Model 4
Coefficients
t-ratio
0.075
2.58
0.003
0.40
0.028
2.50
Coefficients
t-ratio
0.075
2.58
0.003
0.34
Coefficients
t-ratio
0.076
2.55
0.004
0.50
Coefficients
t-ratio
0.095
3.00
-0.025
-3.13
***
***
**
0.076
2.45
**
0.052
1.71
1-ASSETSH
*
NONUTIL
KZL
NSEG*KZH
NSEG*KZL
-0.023
-1.58
0.059
3.78
-0.003
-0.18
-0.064
-4.39
***
-0.029
-2.14
0.060
4.03
SALEH*KZL
**
***
-0.035
-2.53
0.057
3.92
ASSETSH*KZL
NONUTIL*KZH
NONUTIL*KZL
15.94
0.000
0.904
1434
***
***
0.021
0.54
-0.185
-4.58
ASSETSH*KZH
F statistics
P value
Adj. R-squared
No. of Obs.
**
0.096
2.01
-0.027
-2.26
0.025
1.93
***
0.018
0.45
-0.191
-4.77
SALEH*KZH
***
**
1-SALEH
KZH
***
23.43
0.000
0.904
1432
- 33 -
22.10
0.000
0.905
1411
***
0.027
0.46
-0.073
-1.39
3.23
0.073
0.916
1343
**
**
*
Table 6: Robustness check: Sub-sample of parents which change diversification
The table restricts the sample to parent firms which change their degree of diversification. The
dependent variable is TFP, defined as the residual from estimating equation (2). KZH (KZL) is a dummy
variable that takes 1 when the firm’s KZ index belongs to the top (bottom) 1/3 of the variable for each
year and 0 otherwise. All models include operating firm fixed effects. Absolute t-statistics are given
below the coefficient estimates. *, **, and *** indicate significance at the 10%, 5%, and 1% levels,
respectively.
Variables
OPSIZE
PSIZE
NSEG
Model 1
Model 2
Model 3
Model 4
Coefficients
t-ratio
0.069
2.30
0.003
0.32
0.030
2.62
Coefficients
t-ratio
0.069
2.31
0.002
0.24
Coefficients
t-ratio
0.070
2.27
0.003
0.36
Coefficients
t-ratio
0.092
2.81
-0.025
-3.15
**
**
**
0.084
2.61
***
0.061
1.93
1-ASSETSH
*
NONUTIL
KZL
NSEG*KZH
NSEG*KZL
-0.025
-1.66
0.072
4.39
0.000
-0.02
-0.073
-4.81
*
***
-0.032
-2.27
0.073
4.65
SALEH*KZL
**
***
-0.039
-2.69
0.070
4.54
ASSETSH*KZL
NONUTIL*KZH
NONUTIL*KZL
20.55
0.000
0.903
1358
***
**
**
**
***
0.027
0.65
-0.210
-5.03
ASSETSH*KZH
F statistics
P value
Adj. R-squared
No. of Obs.
***
0.102
2.07
-0.029
-2.25
0.033
2.41
***
0.025
0.58
-0.217
-5.22
SALEH*KZH
***
***
1-SALEH
KZH
***
29.22
0.000
0.904
1356
- 34 -
27.38
0.000
0.904
1335
***
0.036
0.61
-0.090
-1.66
4.91
0.027
0.916
1269
*
Table 7: Robustness check: Impact of regulation
The table examines the impact of regulation. The dependent variable is TFP, defined as the residual from
estimating equation (2). KZH (KZL) is a dummy variable that takes 1 when the firm’s KZ index belongs
to the top (bottom) 1/3 of the variable for each year and 0 otherwise. Reg96 is a dummy variable that
takes the value of 1 during the period from 1996 to 2003, and 0 otherwise. All models include operating
firm fixed effects. Absolute t-statistics are given below the coefficient estimates. *, **, and *** indicate
significance at the 10%, 5%, and 1% levels, respectively.
Variables
OPSIZE
PSIZE
NSEG
Model 1
Model 2
Model 3
Model 4
Coefficients
t-ratio
0.075
2.52
0.004
0.52
0.037
2.58
Coefficients
t-ratio
0.073
2.44
0.002
0.25
Coefficients
t-ratio
0.075
2.45
0.005
0.58
Coefficients
t-ratio
0.092
2.78
-0.026
-3.15
**
**
**
0.071
1.86
*
0.086
2.16
1-ASSETSH
**
NONUTIL
KZL
NSEG*KZH
NSEG*KZL
-0.023
-1.56
0.059
3.80
-0.002
-0.16
-0.064
-4.38
***
-0.029
-2.14
0.060
4.02
(1-SALEH)*KZL
**
***
-0.033
-2.45
0.057
3.94
(1-ASSETSH)*KZL
NONUTIL*KZH
NONUTIL*KZL
NSEG*Reg96
0.009
0.90
-0.014
-1.15
***
***
0.022
0.56
-0.185
-4.58
(1-ASSETSH)*KZH
Reg96
**
0.066
0.06
-0.027
-2.24
0.025
1.94
***
0.018
0.45
-0.191
-4.77
(1-SALEH)*KZH
0.002
0.16
0.012
1.29
***
0.021
0.37
-0.077
-1.45
-0.002
-0.26
0.003
0.10
(1-SALEH)*Reg96
-0.051
-1.53
(1-ASSETSH)*Reg96
0.037
NONUTIL*Reg96
F statistics
P value
16.00
0.000
23.41
0.000
22.29
0.000
0.94
3.12
0.078
Adj. R-squared
0.904
0.904
0.905
0.916
1434
1432
1411
1343
No. of Obs.
***
***
1-SALEH
KZH
***
- 35 -
**
*